Do Computers Lie?
Machine Bias
The Problem
The Solution
Libraries
Ignorance is Bliss - but not always!
Reproducibility, Explainability and Bias
Model Bias & Model Fairness
Model Interpretability / Explaining ML Models
Recommendations from these Insights
Media Coverage about Bias / IML
Undoing Bias / Improving IML - Papers and Attempts

Do Computers Lie?

We have been constantly told this statement “Computers don’t lie”. Yes in fact Computers don’t lie, but neither does it speak the truth. A computer does what its Master programs it to do. Similarly, A model wouldn’t lie unless the Machine Learning Engineer doesn’t want it to lie.

Machine Bias

There was a nice episode of the podcast You are not so smart came out last year. This is an excerpt from it:

“I want a machine-learning algorithm to learn what tumors looked like in the past, and I want it to become biased toward selecting those kind of tumors in the future,” explains philosopher Shannon Vallor at Santa Clara University. “But I don’t want a machine-learning algorithm to learn what successful engineers and doctors looked like in the past and then become biased toward selecting those kinds of people when sorting and ranking resumes.”

The Problem

Machine Bias can occur due to a lot of factors but a few to name is:

Biased Training Dataset
Bias Variable in the Feature Space
BlackBox Modelling of not understanding what’s going on with the Model

Below is an example of how Google Translate, when translated the following text to a Gender-neutral langauge and back to English - applies its bias (primarily due to the nature of biased Training Dataset)

img

The Solution

The first step of finding solution to any problem is accepting The Problem exists. Let’s accept that fact and see how to use Kaggle Survey results and help the community tackle Machine Bias.

Libraries

suppressPackageStartupMessages(library(tidyverse)) 
suppressPackageStartupMessages(library(highcharter))
suppressPackageStartupMessages(library(DataExplorer))
suppressPackageStartupMessages(library(scales))
suppressPackageStartupMessages(library(cowplot))
suppressPackageStartupMessages(library(viridis))
suppressPackageStartupMessages(library(wordcloud))
suppressPackageStartupMessages(library(tidytext))
suppressPackageStartupMessages(library(RColorBrewer))

pal <- brewer.pal(9,"BuGn")


plotting_missing <- function(df){

  #based on erikbruin's code snippet
    
NAcol <- which(colSums(is.na(df)) > 0)
NAcount <- sort(colSums(sapply(df[NAcol], is.na)), decreasing = TRUE)
NADF <- data.frame(variable=names(NAcount), missing=NAcount)
NADF$PctMissing <- round(((NADF$missing/nrow(df))*100),1)
NADF %>%
    ggplot(aes(x=reorder(variable, PctMissing), y=PctMissing)) +
    geom_bar(stat='identity', fill='red') + coord_flip(y=c(0,110)) +
    labs(x="", y="Percent missing") +
    geom_text(aes(label=paste0(NADF$PctMissing, "%"), hjust=-0.1))
}

survey <- read_csv("input/multipleChoiceResponses.csv", skip = 1, col_types = cols())

Ignorance is Bliss - but not always!

p1  + theme_minimal() + theme(axis.text = element_text(size = 6)) + labs(title = "Ignorance beyond DS")

The above plot is to demonstrate how much these questions that are about Model Fairness / Bias, have been ignored.

While asking about Salary made 15% of respondents to not answer, Questions about Reproducibility, Explainability and Bias made 37% of respondents to skip answering. The salary question comparsion is here to show relatively worse questions like this are approached.

Reproducibility, Explainability and Bias

survey %>% select(contains("How do you perceive the importance of the following topics?")) %>% 
  gather() %>% 
  mutate(key = str_replace(key,"-","\n")) %>% 
  mutate(key = str_replace(key,"How do you perceive the importance of the following topics?",""),
         key = str_replace(key, regex("\\?"),""), 
         key = str_replace(key, regex("\\-|\\:"),"")) %>% 
  group_by(key) %>% 
  count(value) %>% 
  drop_na() %>% 
  mutate(n = n / sum(n)) %>% 
  ggplot() + geom_col(aes(value,n, fill = key), stat = "identity", show.legend = FALSE) +
   geom_label(aes(x = value, y = n - 0.05, label = percent(n)),
           # hjust=0, vjust=0, size = 4, colour = 'black',
            fontface = 'bold') +
  scale_fill_viridis(discrete = T, option = "E") +
  facet_wrap(~key) +
  scale_y_continuous(labels = percent_format()) +
  theme_minimal() + 
  theme(axis.text = element_text(angle = 45, size = 6)) +
  labs(title = "Perception on Reproducibility, Explainability and Model Bias ",
       subtitle = "Percentage",
       x = "Selected Options",
       y = "Percentage of Respondents (other than NAs)")

To get a better perspective of the volume of the respondents, below is the same plot as above but with absolute numbers of respondents and their options.

survey %>% select(contains("How do you perceive the importance of the following topics?")) %>% 
  gather() %>% 
  mutate(key = str_replace(key,"-","\n")) %>% 
  mutate(key = str_replace(key,"How do you perceive the importance of the following topics?",""),
         key = str_replace(key, regex("\\?"),""), 
         key = str_replace(key, regex("\\-|\\:"),"")) %>% 
  group_by(key) %>% 
  count(value) %>% 
  drop_na()  %>% 
  
  ggplot() + geom_col(aes(value,n, fill = key), stat = "identity", show.legend = FALSE) +
   geom_label(aes(x = value, y = n + 20, label = n),
           # hjust=0, vjust=0, size = 4, colour = 'black',
            fontface = 'bold') +
  
  scale_fill_viridis(discrete = T, option = "E") +
  facet_wrap(~key) +
  #scale_y_continuous(labels = percent_format()) +
  theme_minimal() + 
  theme(axis.text = element_text(angle = 45, size = 6)) +
  labs(title = "Perception on Reproducibility, Explainability and Model Bias ",
       subtitle = "Absolute Numbers",
       x = "Selected Options",
       y = "Percentage of Respondents (other than NAs)")

Fairness and Bias:

Only close to half (57.4%) of the respondenrts who chose to answer consider Fairness and Bias in ML Algorithm is a Very important.
This is the lowest Very important sentiment echoed by the community of all the 3 questions.
3.6% of those who chose to respondent perceive this is Not at all important, which is the highest Not at all important feeling expressed of all 3 questions.

survey %>% select(contains("How do you perceive the importance of the following topics?")) %>% 
  gather() %>% 
  mutate(key = str_replace(key,"How do you perceive the importance of the following topics?",""),
         key = str_replace(key, regex("\\?"),""), 
         key = str_replace(key, regex("\\-"),"")) %>% 
  group_by(key) %>% 
  count(value) %>% 
  mutate(n = percent(n / sum(n)))  %>% 
  spread(value,n) %>% 
  knitr::kable()

key	No opinion; I do not know	Not at all important	Slightly important	Very important
Being able to explain ML model outputs and/or predictions	2.9%	1.6%	17.0%	41.1%	37.4%
Fairness and bias in ML algorithms:	5.4%	2.3%	19.0%	36.0%	37.4%
Reproducibility in data science	3.8%	1.0%	14.9%	42.9%	37.4%

Model Bias & Model Fairness

Who are they?

They refer to those beings who think Fairness and Bias in ML Alggorithm are Very Important in Machine Learning.

they <- survey %>% filter(`How do you perceive the importance of the following topics? - Fairness and bias in ML algorithms:` == "Very important")


not_they <- survey %>% filter(`How do you perceive the importance of the following topics? - Fairness and bias in ML algorithms:` != "Very important")

Gender

they %>% group_by(`What is your gender? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Gender" = `What is your gender? - Selected Choice`) %>% 
  mutate(n = n / sum(n),
         perc = percent(n)) %>% 
   ggplot() + geom_col(aes(Gender,n, fill = Gender), stat = "identity", show.legend = FALSE) +
   geom_label(aes(x = Gender, y = n - 0.05, label = percent(n)),
           # hjust=0, vjust=0, size = 4, colour = 'black',
            fontface = 'bold') +
    scale_fill_viridis(discrete = T, option = "E") +

  scale_y_continuous(labels = percent_format()) +
  theme_minimal() + 
  theme(axis.text = element_text(angle = 45, size = 6)) +
  labs(title = "They",
       subtitle = "Perception on Fairness and Model Bias ",
       x = "Gender",
       y = "Percentage of Respondents (other than NAs)") -> p1



not_they %>% group_by(`What is your gender? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Gender" = `What is your gender? - Selected Choice`) %>% 
  mutate(n = n / sum(n),
         perc = percent(n)) %>% 
    ggplot() + geom_col(aes(Gender,n, fill = Gender), stat = "identity", show.legend = FALSE) +
   geom_label(aes(x = Gender, y = n - 0.05, label = percent(n)),
           # hjust=0, vjust=0, size = 4, colour = 'black',
            fontface = 'bold') +
    scale_fill_viridis(discrete = T, option = "E") +

  scale_y_continuous(labels = percent_format()) +
  theme_minimal() + 
  theme(axis.text = element_text(angle = 45, size = 6)) +
  labs(title = "Not They",
       subtitle = "Perception on Fairness and Model Bias ",
       x = "Gender",
       y = "Percentage of Respondents (other than NAs)") -> p2


cowplot::plot_grid(p1,p2)

Let us a create a new KPI index called They - Not They Ratio to give a different perspective to this comparison.

they %>% filter(!`What is your gender? - Selected Choice` %in% c("Other")) %>% 
  group_by(`What is your gender? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Degree" = `What is your gender? - Selected Choice`,
         "They" = n) %>%  
  bind_cols(
not_they %>% filter(!`What is your gender? - Selected Choice` %in% c("Other")) %>% 
  group_by(`What is your gender? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Degree1" = `What is your gender? - Selected Choice`,
         "Not They" = n) %>% 
  select(-Degree1)) %>% 
  mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("column",hcaes("Degree","T_NT_Ratio")) %>% 
  hc_title(text = "Gender-wise preference - Model Bias and Fairness  ") %>% 
  hc_add_theme(hc_theme_538())

There is a difference of 5.1 PP Female Percentage difference between those who perceive Model Fariness & Bias in ML is Very Important and Others.
While this could be seen as that Female Gender usually gets affected by these Biases, It’s also important to realize that Male Gender (Kaggler’s) don’t echo similar sentiment as their female counterpart. After all, A healthy model is what we all want, don’t we?
Using the index that we created T_NT_Ratio, we can see that Female gender and those who selected other options other than Male are above on top of Male Kagglers in their perception about Model Bias and Fairness

Age

they %>% group_by(`What is your age (# years)?`) %>% count() %>% ungroup() %>% 
  rename("Age" = `What is your age (# years)?`) %>% 
  mutate(n = n / sum(n),
         perc = percent(n)) %>% 
   ggplot() + geom_col(aes(Age,n, fill = Age), stat = "identity", show.legend = FALSE) +
   geom_label(aes(x = Age, y = n - 0.05, label = percent(n)),
           # hjust=0, vjust=0, size = 4, colour = 'black',
            fontface = 'bold') +
    scale_fill_viridis(discrete = T, option = "D") +
  scale_y_continuous(labels = percent_format()) +
  theme_minimal() + 
  theme(axis.text = element_text(angle = 45, size = 6)) +
  labs(title = "They",
       subtitle = "Perception on Fairness and Model Bias ",
       x = "Age",
       y = "Percentage of Respondents (other than NAs)") -> p1



not_they %>% group_by(`What is your age (# years)?`) %>% count() %>% ungroup() %>% 
  rename("Age" = `What is your age (# years)?`) %>% 
  mutate(n = n / sum(n),
         perc = percent(n)) %>% 
    ggplot() + geom_col(aes(Age,n, fill = Age), stat = "identity", show.legend = FALSE) +
   geom_label(aes(x = Age, y = n - 0.05, label = percent(n)),
           # hjust=0, vjust=0, size = 4, colour = 'black',
            fontface = 'bold') +
    scale_fill_viridis(discrete = T, option = "D") +
  scale_y_continuous(labels = percent_format()) +
  theme_minimal() + 
  theme(axis.text = element_text(angle = 45, size = 6)) +
  labs(title = "Not They",
       subtitle = "Perception on Fairness and Model Bias ",
       x = "Age",
       y = "Percentage of Respondents (other than NAs)") -> p2


cowplot::plot_grid(p1,p2)

Age doesn’t seem to give anything straightway, which probably could be due to a lot of different age brackets. Let us try a bit of engineering to club them into two groups < 30 and > 30.

they %>% 
  mutate(age_grp = ifelse(parse_number(`What is your age (# years)?`) < 30,
                          "Less than 30",
                          "30+")) %>% 
  group_by(age_grp) %>% count() %>% ungroup() %>% 
  rename("Age" = age_grp) %>% 
  mutate(n = n / sum(n),
         perc = percent(n)) %>% 
   ggplot() + geom_col(aes(Age,n, fill = Age), stat = "identity", show.legend = FALSE) +
   geom_label(aes(x = Age, y = n - 0.05, label = percent(n)),
           # hjust=0, vjust=0, size = 4, colour = 'black',
            fontface = 'bold') +
    scale_fill_viridis(discrete = T, option = "E") +
  scale_y_continuous(labels = percent_format()) +
  theme_minimal() + 
  theme(axis.text = element_text(angle = 45, size = 6)) +
  labs(title = "They",
       subtitle = "Perception on Fairness and Model Bias ",
       x = "Age",
       y = "Percentage of Respondents (other than NAs)") -> p1



not_they %>% 
   mutate(age_grp = ifelse(parse_number(`What is your age (# years)?`) < 30,
                          "Less than 30",
                          "30+")) %>% 
  group_by(age_grp) %>% count() %>% ungroup() %>% 
  rename("Age" = age_grp) %>% 
  mutate(n = n / sum(n),
         perc = percent(n)) %>% 
    ggplot() + geom_col(aes(Age,n, fill = Age), stat = "identity", show.legend = FALSE) +
   geom_label(aes(x = Age, y = n - 0.05, label = percent(n)),
           # hjust=0, vjust=0, size = 4, colour = 'black',
            fontface = 'bold') +
    scale_fill_viridis(discrete = T, option = "E") +
  scale_y_continuous(labels = percent_format()) +
  theme_minimal() + 
  theme(axis.text = element_text(angle = 45, size = 6)) +
  labs(title = "Not They",
       subtitle = "Perception on Fairness and Model Bias ",
       x = "Age",
       y = "Percentage of Respondents (other than NAs)") -> p2


cowplot::plot_grid(p1,p2)

This plot helps us say that the younger ones need to be updated with the implications of Model Bias and Fairness more than their older counterparts. That leads us to another important section of what they do.

Student vs Professionals

they %>% 
  mutate(title = ifelse(`Select the title most similar to your current role (or most recent title if retired): - Selected Choice` == "Student",
                          "Student",
                          "Professional")) %>% 
  group_by(title) %>% count() %>% ungroup() %>% 
  rename("Title" = title) %>% 
  mutate(n = n / sum(n),
         perc = percent(n)) %>% 
   ggplot() + geom_col(aes(Title,n, fill = Title), stat = "identity", show.legend = FALSE) +
   geom_label(aes(x = Title, y = n - 0.05, label = percent(n)),
           # hjust=0, vjust=0, size = 4, colour = 'black',
            fontface = 'bold') +
    scale_fill_viridis(discrete = T, option = "C") +
  scale_y_continuous(labels = percent_format()) +
  theme_minimal() + 
  theme(axis.text = element_text(angle = 45, size = 6)) +
  labs(title = "They",
       subtitle = "Perception on Fairness and Model Bias ",
       x = "Title",
       y = "Percentage of Respondents (other than NAs)") -> p1


not_they %>% 
  mutate(title = ifelse(`Select the title most similar to your current role (or most recent title if retired): - Selected Choice` == "Student",
                          "Student",
                          "Professional")) %>% 
  group_by(title) %>% count() %>% ungroup() %>% 
  rename("Title" = title) %>% 
  mutate(n = n / sum(n),
         perc = percent(n)) %>% 
    ggplot() + geom_col(aes(Title,n, fill = Title), stat = "identity", show.legend = FALSE) +
   geom_label(aes(x = Title, y = n - 0.05, label = percent(n)),
           # hjust=0, vjust=0, size = 4, colour = 'black',
            fontface = 'bold') +
  scale_fill_viridis(discrete = T, option = "C") +
  scale_y_continuous(labels = percent_format()) +
  theme_minimal() + 
  theme(axis.text = element_text(angle = 45, size = 6)) +
  labs(title = "Not They",
       subtitle = "Perception on Fairness and Model Bias ",
       x = "Title",
       y = "Percentage of Respondents (other than NAs)") -> p2


cowplot::plot_grid(p1,p2)

As with the previouus insight, once again it comes that Students need to be educated with the concepts of Model Bias and Fairness since there is ~4 PP difference between Students in They vs Not They.

Undergraduate Majors

they %>% 
  mutate(UG = ifelse(`Which best describes your undergraduate major? - Selected Choice` %in% c("Computer science (software engineering, etc.)","Information technology, networking, or system administration"),
                          "CS",
                          "Non_CS")) %>% 
  group_by(UG) %>% count() %>% ungroup() %>% 
  rename("UG" = UG) %>% 
  mutate(n = n / sum(n),
         perc = percent(n)) %>% 
   ggplot() + geom_col(aes(UG,n, fill = UG), stat = "identity", show.legend = FALSE) +
   geom_label(aes(x = UG, y = n - 0.05, label = percent(n)),
           # hjust=0, vjust=0, size = 4, colour = 'black',
            fontface = 'bold') +
  scale_fill_viridis(discrete = T, option = "D") +
  scale_y_continuous(labels = percent_format()) +
  theme_minimal() + 
  theme(axis.text = element_text(angle = 45, size = 6)) +
  labs(title = "They",
       subtitle = "Perception on Fairness and Model Bias ",
       x = "Title",
       y = "Percentage of Respondents (other than NAs)") -> p1


not_they %>% 
   mutate(UG = ifelse(`Which best describes your undergraduate major? - Selected Choice` %in% c("Computer science (software engineering, etc.)","Information technology, networking, or system administration"),
                          "CS",
                          "Non_CS")) %>% 
  group_by(UG) %>% count() %>% ungroup() %>% 
  rename("UG" = UG) %>% 
  mutate(n = n / sum(n),
         perc = percent(n)) %>% 
   ggplot() + geom_col(aes(UG,n, fill = UG), stat = "identity", show.legend = FALSE) +
   geom_label(aes(x = UG, y = n - 0.05, label = percent(n)),
           # hjust=0, vjust=0, size = 4, colour = 'black',
            fontface = 'bold') +
  scale_fill_viridis(discrete = T, option = "D") +
  scale_y_continuous(labels = percent_format()) +
  theme_minimal() + 
  theme(axis.text = element_text(angle = 45, size = 6)) +
  labs(title = "Not They",
       subtitle = "Perception on Fairness and Model Bias ",
       x = "Title",
       y = "Percentage of Respondents (other than NAs)") -> p2


cowplot::plot_grid(p1,p2)

Comptuer Science / IT Engineers have a difference of ~ 3.4PP between They and Not They, which shows that’s needs more care in the mindset change than Non-CS (even though that’s also required) but at least this can help prioritise where to start with any campaigning

R vs Python vs more

they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`What specific programming language do you use most often? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Lang" = `What specific programming language do you use most often? - Selected Choice`,
         "They" = n) %>%  
  bind_cols(
not_they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`What specific programming language do you use most often? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Lang1" = `What specific programming language do you use most often? - Selected Choice`,
         "Not They" = n) %>% 
  select(-Lang1)) %>% 
  mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("bar",hcaes("Lang","T_NT_Ratio")) %>% 
  hc_title(text = "Language ordered by They-NotThey Ratio") %>% 
  hc_add_theme(hc_theme_538())

R users tend to be perceive Model Bias and Fairness as Very Important than their Python Counterparts.

All Countries

they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`In which country do you currently reside?`) %>% count() %>% ungroup() %>% 
  rename("Country" = `In which country do you currently reside?`,
         "They" = n) %>%  
  bind_cols(
not_they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`In which country do you currently reside?`) %>% count() %>% ungroup() %>% 
  rename("Country1" = `In which country do you currently reside?`,
         "Not They" = n) %>% 
  select(-Country1)) %>% 
  mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("line",hcaes("Country","T_NT_Ratio")) %>% 
  hc_title(text = "They vs Not They Ratio - Country-wise") %>% 
    hc_add_theme(hc_theme_538())

Countries with at least 100 respondents

they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`In which country do you currently reside?`) %>% count() %>% ungroup() %>% 
  rename("Country" = `In which country do you currently reside?`,
         "They" = n) %>%  
  bind_cols(
not_they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`In which country do you currently reside?`) %>% count() %>% ungroup() %>% 
  rename("Country1" = `In which country do you currently reside?`,
         "Not They" = n) %>% 
  select(-Country1)) %>% 
  filter((They + `Not They`) > 100) %>% 
  mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("line",hcaes("Country","T_NT_Ratio")) %>% 
  hc_title(text = "They vs Not They Ratio - Country-wise > 100 respondents") %>% 
    hc_add_theme(hc_theme_538())

Chile leads the pack if all the countries are considered otherwise if only the countries with more than 100 respondents are selected, South Africa followed by Nigeria are leading the pack of a healthy They to Not They ratio.

Media Sources

they %>% 
  dplyr::select(contains("your favorite media sources that report on data science topics")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:22, key = "questions", value = "media") %>% 
 
  #tidyr::replace_na() %>% 
  group_by(media) %>% 
  count() %>% 
  rename("They" = n) %>% 
bind_cols(
 

not_they %>% 
  dplyr::select(contains("your favorite media sources that report on data science topics")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:22, key = "questions", value = "media") %>% 
 
  group_by(media) %>% 
  count() %>% 
  rename("Not They" = n)
) %>% 
  dplyr::select(-media1) %>% 
mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  drop_na() %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("line",hcaes("media","T_NT_Ratio")) %>% 
  hc_title(text = " Model Bias and Fairness - Media Sources - They vs Not They Ratio") %>% 
    hc_add_theme(hc_theme_538())

The above plot makes it very clear that Podcasts like Partiall Derivative, Data Skpetic and Linear Digression are places where Kagglers who strongly believe that ML Bias and Fairness get their news from.
Nate Silver’s 538 makes an entry in the top 5 media sources arranged by T_NT_Ratio KPI
Kaggle Forums seemed to need to make more work in terms of encouraging or initiating discussions about ML Bias and Fairness

Work Industry

they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`In what industry is your current employer/contract (or your most recent employer if retired)? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Industry" = `In what industry is your current employer/contract (or your most recent employer if retired)? - Selected Choice`,
         "They" = n) %>%  
  bind_cols(
not_they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`In what industry is your current employer/contract (or your most recent employer if retired)? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Industry1" = `In what industry is your current employer/contract (or your most recent employer if retired)? - Selected Choice`,
         "Not They" = n) %>% 
  select(-Industry1)) %>% 
  #filter((They + `Not They`) > 100) %>% 
  mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  filter(!Industry %in% c("3","Other")) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("line",hcaes("Industry","T_NT_Ratio")) %>% 
  hc_title(text = "They vs Not They Ratio - Industry-wise") %>% 
    hc_add_theme(hc_theme_538())

Kagglers in Industries like Non_Profit/Service and Government/Public Service have been better perception about the importance of Model Fairness and Bias.
The above plot once again empahsis the importance of Model Bias and Fairness education among students.
It’s also unhealthy to see places like Military and Internet-based Services falling behind as those are the places where the model evaluation is crucial and can have serious outcomes.

Data Scientist?

they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`Do you consider yourself to be a data scientist?`) %>% count() %>% ungroup() %>% 
  rename("DS" = `Do you consider yourself to be a data scientist?`,
         "They" = n) %>%  
  bind_cols(
not_they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`Do you consider yourself to be a data scientist?`) %>% count() %>% ungroup() %>% 
  rename("DS1" = `Do you consider yourself to be a data scientist?`,
         "Not They" = n)# %>% 
 # select(-DS1)
) %>% 
  #filter((They + `Not They`) > 100) %>% 
  mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  filter(!DS %in% c("3","Other")) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("line",hcaes("DS","T_NT_Ratio")) %>% 
  hc_title(text = "They vs Not They Ratio - Based on if they consider themsevles a Data Scientist") %>% 
    hc_add_theme(hc_theme_538())

Those who consider themselves to be Definitely a data scientist are more likely also to believe about Model Bias and Fairness with those who do not consider themselves a data scientists making the last end of the spectrum with lowest They vs Not They Ratio!

Type of Data

they %>% 
  dplyr::select(contains("types of data")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:11, key = "questions", value = "DataType") %>% 
  #tidyr::replace_na() %>% 
  group_by(DataType) %>% 
  count() %>% 
  rename("They" = n) %>% 
bind_cols(
  not_they %>% 
  dplyr::select(contains("types of data")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:11, key = "questions", value = "DataType") %>% 
  #tidyr::replace_na() %>% 
  group_by(DataType) %>% 
  count() %>% 
  rename("Not They" = n)
) %>% 
  dplyr::select(-DataType1) %>% 
mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("line",hcaes("DataType","T_NT_Ratio")) %>% 
  hc_title(text = "Type of Data - They vs Not They Ratio ") %>% 
    hc_add_theme(hc_theme_538())

Kagglers who use Genetic Data lead the table of giving importance to Model Bias
Kagglers who use Image Data and Video Data are least bothered

MOOC - Online course Platform

they %>% 
  dplyr::select(contains("online platforms have you begun or completed data science courses")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:13, key = "questions", value = "DataType") %>% 
 
  #tidyr::replace_na() %>% 
  group_by(DataType) %>% 
  count() %>% 
  rename("They" = n) %>% 
bind_cols(
 

not_they %>% 
  dplyr::select(contains("online platforms have you begun or completed data science courses")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:13, key = "questions", value = "DataType") %>% 
 
  group_by(DataType) %>% 
  count() %>% 
  rename("Not They" = n)
) %>% 
  dplyr::select(-DataType1) %>% 
mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("column",hcaes("DataType","T_NT_Ratio", color = "T_NT_Ratio")) %>% 
   hc_title(text = " MOOC - Online course platform - They vs Not They Ratio") %>% 
   hc_add_theme(hc_theme_538())

Ironically, Kaggle Learn combining with Coursera and Udemy top the places from where people who don’t think Model Bias is very important learn Data Science / Machine Learning
Kagglers who learn DS from Google Developers, Theschool.ai and Fast.ai are the ones who lead the thought process of thinking that Model Bias is very important

Percentage of Data Projects

survey %>% 
  select(`Approximately what percent of your data projects involved exploring unfair bias in the dataset and/or algorithm?`) %>% 
  drop_na() %>% 
  group_by(`Approximately what percent of your data projects involved exploring unfair bias in the dataset and/or algorithm?`) %>% 
  rename("Percentage" = `Approximately what percent of your data projects involved exploring unfair bias in the dataset and/or algorithm?`) %>% 
  count() %>% 
  hchart("area",hcaes("Percentage","n")) %>% 
  hc_title(text = "What percent of your data projects involved exploring unfair bias ") %>% 
  hc_add_theme(hc_theme_538())

As you can see above, There is a very very minimal demand in the workplace to explore unfair ML bias.
In fact, only ~ 1.2K respondents of the total ~13K respondents who respondended to this question, have answered to have a need of more than 50% to explore unfair bias in their data projects. (that’s less than 10%)

Metrics for Model Success

survey %>% 
  dplyr::select(contains("not your models were successful")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:5, key = "questions", value = "Metric") %>% 
  #tidyr::replace_na() %>% 
  group_by(Metric) %>% 
  count()  %>% 

  drop_na() %>% 

  hchart("column",hcaes("Metric","n")) %>% 
  hc_title(text = " Metrics used in organizations to determine whether or not your models were successful") %>% 
    hc_add_theme(hc_theme_darkunica())

While it’s obvious still a fact that Most people would say that usual Model Accuracy metrics are considered to determine whether a model is successful or not
Revenue / Business Goals has almost 2X more votes than Metri s that consider unfair Bias

Difficulty in implementing a Fair / Bias-free ML Model

survey %>% 
  dplyr::select(contains("difficult about ensuring that your algorithms are fair and unbiased")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:6, key = "questions", value = "Difficulty") %>% 
  #tidyr::replace_na() %>% 
  group_by(Difficulty) %>% 
  count()  %>% 

  drop_na() %>% 

  hchart("bar",hcaes("Difficulty","n")) %>% 
  hc_title(text = "Most difficult about ensuring that your algorithms are fair and unbiased?") %>% 
    hc_add_theme(hc_theme_darkunica())

Collecting relevant Data without Bias is one of the most prominent difficulty in building a Bias free and Fair Model
That is followed by, Picking the right Evaluation Metric for the model’s ML Bias and Identifying the group that’s been unfairly targetted in the data so that can be used to unbias / nullify the effect

Model Interpretability / Explaining ML Models

Who are they?

They refer to those beings who think Being able to explain ML model outputs and/or predictions are Very Important in Machine Learning and Not They are those who think otherwise including - Somewhat Important, Not at all important and similar.

they <- survey %>% filter(`How do you perceive the importance of the following topics? - Being able to explain ML model outputs and/or predictions` == "Very important")


not_they <- survey %>% filter(`How do you perceive the importance of the following topics? - Being able to explain ML model outputs and/or predictions` != "Very important")

Gender

they %>% filter(!`What is your gender? - Selected Choice` %in% c("Other")) %>% 
  group_by(`What is your gender? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Degree" = `What is your gender? - Selected Choice`,
         "They" = n) %>%  
  bind_cols(
not_they %>% filter(!`What is your gender? - Selected Choice` %in% c("Other")) %>% 
  group_by(`What is your gender? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Degree1" = `What is your gender? - Selected Choice`,
         "Not They" = n) %>% 
  select(-Degree1)) %>% 
  mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("column",hcaes("Degree","T_NT_Ratio")) %>% 
  hc_title(text = "Gender-wise preference - Interpretable Machine Learning  ") %>% 
  hc_add_theme(hc_theme_538())

Age

they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`What is your age (# years)?`) %>% count() %>% ungroup() %>% 
  rename("Age" = `What is your age (# years)?`,
         "They" = n) %>%  
  bind_cols(
not_they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`What is your age (# years)?`) %>% count() %>% ungroup() %>% 
  rename("Age1" = `What is your age (# years)?`,
         "Not They" = n) %>% 
  select(-Age1)) %>% 
  mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("area",hcaes("Age","T_NT_Ratio")) %>% 
  hc_title(text = "Age Range - Interpretable Machine Learning  ") %>% 
  hc_add_theme(hc_theme_538())

The cohort between Age group 18 - 29 are almost half of the index value of the cohort in 60-69 which shows how age and experience play a vital role in their perception about Interpretable Machine Learning

R vs Python vs more

they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`What specific programming language do you use most often? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Lang" = `What specific programming language do you use most often? - Selected Choice`,
         "They" = n) %>%  
  bind_cols(
not_they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`What specific programming language do you use most often? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Lang1" = `What specific programming language do you use most often? - Selected Choice`,
         "Not They" = n) %>% 
  select(-Lang1)) %>% 
  mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("bar",hcaes("Lang","T_NT_Ratio")) %>% 
  hc_title(text = "Language - Interpretable Machine Learning  ") %>% 
  hc_add_theme(hc_theme_538())

Unexpectedly, SAS/STATA tops the index of They-Not_they ratio followed by R and MATLAB - making these three language users on Kaggle believe Model Intepretability is very important.
Python users need to be made aware of IML as Python still catches up behind SQL and Julia.

Degree - Education

they %>% filter(!`Which best describes your undergraduate major? - Selected Choice` %in% c("Other")) %>% 
  group_by(`Which best describes your undergraduate major? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Degree" = `Which best describes your undergraduate major? - Selected Choice`,
         "They" = n) %>%  
  bind_cols(
not_they %>% filter(!`Which best describes your undergraduate major? - Selected Choice` %in% c("Other")) %>% 
  group_by(`Which best describes your undergraduate major? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Degree1" = `Which best describes your undergraduate major? - Selected Choice`,
         "Not They" = n) %>% 
  select(-Degree1)) %>% 
  mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("column",hcaes("Degree","T_NT_Ratio")) %>% 
  hc_title(text = "Degree (Education) -  Interpretable Machine Learning ") %>% 
  hc_add_theme(hc_theme_538())

Kagglers with Humanities background strongly believe in IML (Interpretable Machine Learning) with a TNT index of 2.28, followed by Engineerg (Non-CS) and Mathematics/Statistitcs
Ironically, Computer Science Engineers are no where near the top with the index of 1.1764

Work Industry

they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`In what industry is your current employer/contract (or your most recent employer if retired)? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Industry" = `In what industry is your current employer/contract (or your most recent employer if retired)? - Selected Choice`,
         "They" = n) %>%  
  bind_cols(
not_they %>% #filter(`In which country do you currently reside?` %in% c("India","United States of America")) %>% 
  group_by(`In what industry is your current employer/contract (or your most recent employer if retired)? - Selected Choice`) %>% count() %>% ungroup() %>% 
  rename("Industry1" = `In what industry is your current employer/contract (or your most recent employer if retired)? - Selected Choice`,
         "Not They" = n) %>% 
  select(-Industry1)) %>% 
  #filter((They + `Not They`) > 100) %>% 
  mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  filter(!Industry %in% c("3","Other")) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("line",hcaes("Industry","T_NT_Ratio")) %>% 
  hc_title(text = "IML - They vs Not They Ratio - Industry-wise") %>% 
    hc_add_theme(hc_theme_538())

Kagglers from Insurance and Military/Defense industry background lead the table
As we’ve seen before, Students are no where close to perceive Interpretable Machine Learning is Very important
Industries like Marketing, Online Services are worse than Students in their perception

MOOC - Online course Platform

they %>% 
  dplyr::select(contains("online platforms have you begun or completed data science courses")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:13, key = "questions", value = "DataType") %>% 
 
  #tidyr::replace_na() %>% 
  group_by(DataType) %>% 
  count() %>% 
  rename("They" = n) %>% 
bind_cols(
 

not_they %>% 
  dplyr::select(contains("online platforms have you begun or completed data science courses")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:13, key = "questions", value = "DataType") %>% 
 
  group_by(DataType) %>% 
  count() %>% 
  rename("Not They" = n)
) %>% 
  dplyr::select(-DataType1) %>% 
mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("area",hcaes("DataType","T_NT_Ratio")) %>% 
  hc_title(text = " IML - MOOC - Online course platform - They vs Not They Ratio") %>% 
    hc_add_theme(hc_theme_538())

Kaggle Learn that has an exclusive course on Interpretable Machine Learning hasn’t managed to be on the top, only with the 5 position
Datacamp Kagglers have a strong feeling of Very important about IML
Unlike, Model Bias perception where Fast.ai was on the top, here it’s at the rock-bottom

Media Sources

they %>% 
  dplyr::select(contains("your favorite media sources that report on data science topics")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:22, key = "questions", value = "media") %>% 
 
  #tidyr::replace_na() %>% 
  group_by(media) %>% 
  count() %>% 
  rename("They" = n) %>% 
bind_cols(
 

not_they %>% 
  dplyr::select(contains("your favorite media sources that report on data science topics")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:22, key = "questions", value = "media") %>% 
 
  group_by(media) %>% 
  count() %>% 
  rename("Not They" = n)
) %>% 
  dplyr::select(-media1) %>% 
mutate("T_NT_Ratio" = round(They/`Not They`,3)) %>% 
  drop_na() %>% 
  arrange(desc(T_NT_Ratio)) %>% 
  hchart("line",hcaes("media","T_NT_Ratio")) %>% 
  hc_title(text = " IML - Media Sources - They vs Not They Ratio") %>% 
    hc_add_theme(hc_theme_538())

Kdnuggets, O’reily data newsletter and Kaggle Forum top the list based on the index T_NT_Ratio making them one of the top media sources They get news from the most

Exploring model insights? - Percentage of Data Projects

survey %>% 
  select(`Approximately what percent of your data projects involve exploring model insights?`) %>% 
  drop_na() %>% 
  group_by(`Approximately what percent of your data projects involve exploring model insights?`) %>% 
  rename("Percentage" = `Approximately what percent of your data projects involve exploring model insights?`) %>% 
  count() %>% 
  hchart("area",hcaes("Percentage","n")) %>% 
  hc_title(text = "What percent of your data projects involve exploring model insights?") %>% 
  hc_add_theme(hc_theme_538())

ML models to be black boxes - Difficult to explain

survey %>% 
  select(`Do you consider ML models to be \"black boxes\" with outputs that are difficult or impossible to explain?`) %>% 
  drop_na() %>% 
  group_by(`Do you consider ML models to be \"black boxes\" with outputs that are difficult or impossible to explain?`) %>% 
  rename("Perspective" = `Do you consider ML models to be \"black boxes\" with outputs that are difficult or impossible to explain?`) %>% 
  count() %>% 
  hchart("bar",hcaes("Perspective","n")) %>% 
  hc_title(text = 'ML models to be \"black boxes\" with outputs that are difficult or impossible to explain?') %>% 
  hc_add_theme(hc_theme_538())

The above plot tells us that Most people are confident that they can understand and explain the outputs of Many ML models but not all ML models. In fact, those who feel - Most ML models are Black Boxes are more than those who don’t have any opininon on this matter.

Circumstances where exploring model insights and interpreting model happens

survey %>% 
  dplyr::select(contains("circumstances would you explore model insights and interpret")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:6, key = "questions", value = "circumstances") %>% 
  #tidyr::replace_na() %>% 
  group_by(circumstances) %>% 
  count()  %>% 

  drop_na() %>% 

  hchart("bar",hcaes("circumstances","n")) %>% 
  hc_title(text = "Circumstances where exploring model insights and interpreting model happens") %>% 
    hc_add_theme(hc_theme_darkunica())

* The above plot indicates that this attempt to explore model insights and interpret model happens when The model has been built specificially for that

That’s followed by When first exploring a new ML model or Dataset which tells us an important point that Model Intepretability should be part of AMC or Model Refresh cycle rather than being the first-time duty.

Preferred Methods for Model Interpretability

survey %>% 
  dplyr::select(contains("interpreting decisions that are made by ML models")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:15, key = "questions", value = "methods") %>% 
  #tidyr::replace_na() %>% 
  group_by(methods) %>% 
  count()  %>% 

  drop_na() %>% 

  hchart("bar",hcaes("methods","n")) %>% 
  hc_title(text = " Preferred explaining and/or interpreting decisions that are made by ML models") %>% 
    hc_add_theme(hc_theme_darkunica())

survey %>% 
  select(`What methods do you prefer for explaining and/or interpreting decisions that are made by ML models? (Select all that apply) - Other - Text`) %>% 
  drop_na() %>%
  rename("text" = `What methods do you prefer for explaining and/or interpreting decisions that are made by ML models? (Select all that apply) - Other - Text`) %>% 
  unnest_tokens(word, text)  %>% 
  anti_join(stop_words) %>%
  count(word) %>% 
  arrange(desc(n)) %>% 
  with(wordcloud(word, n, max.words = 100, colors = pal))

## Joining, by = "word"

Plotting Predicted vs Actual Results remain the standalone topper in the list of preferred methods in Interpretable Machine Learning which is kind of an irony because of the fact that it’s an evaluation metric for Model succses rather than Interpreting Machine Learning Model
IML technqiues like LIME, SHAP, ELI5 are way beyond in the list preferences, thus making a point of how much awareness the community has to be edicated to in terms of Interpretable Machine Learning Techniques

Reproducibility

Tools and Methods for ML Reproduction

survey %>% 
  dplyr::select(contains("tools and methods do you use to make your work easy to reproduce")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:11, key = "questions", value = "methods") %>% 
  #tidyr::replace_na() %>% 
  group_by(methods) %>% 
  count()  %>% 

  drop_na() %>% 

  hchart("bar",hcaes("methods","n")) %>% 
  hc_title(text = "Tools and methods do you use to make your work easy to reproduce") %>% 
    hc_add_theme(hc_theme_darkunica())

Documentation and Human-readable codes lead the list of preferred tools and methods to make ML reproducible
Sophisticated ways like Docker or Virtual Box are lot less preferred even ~ 50% lesser than those above easy options
Sharing Code on Github and also code along with Data is another moderately preferred method, that’s less expensive too
Sharing code on Hosted Environment like Kaggle Kernels are in better position than Sophisticated techniques like Docker yet fall a lot behind than easy uploading on Github

Barriers preventing from making easier to reuse and reproducible work

survey %>% 
  dplyr::select(contains(" barriers prevent you from making your work even easier to reuse and reproduce")) %>% 
  dplyr::select(-contains("Text")) %>% 
  gather(1:8, key = "questions", value = "barriers") %>% 
  #tidyr::replace_na() %>% 
  group_by(barriers) %>% 
  count()  %>% 

  drop_na() %>% 

  hchart("bar",hcaes("barriers","n")) %>% 
  hc_title(text = "Barriers preventing from making easier to reuse and reproducible work") %>% 
    hc_add_theme(hc_theme_darkunica())

Producing Reproducible Machine Learning works is considered to be Too Time-consuming which is the most prominent reason why Kagglers don’t prefer doing so.
Another important standing out point followed by the above one is, Not having enough incentive of doing so - reproducible Machine Learning.

Recommendations from these Insights

Exposing College Students and Early in Career Data Scientsits to Model Fairness and Bias Machine Learning
Educating specially Male (and also female) Data scientsts the importance of unBiasing input training data / Model before moving it to production / presentation
Suggesting places like Kaggle and Bootcamps to include Projects/Works about Model Fariness and Bias in their Data Science Curriculum
Metrics to consider a model successful or not should include ML unfair Bias evaluation metrics too.
Improving the avaialbilty of Bias-free / Fair Dataset for Machine Learning. Places like Kaggle Dataset, UCI machine learning should be advised to include Fairness as a label or option.
While it’s good that people who use Genetic Data are concerned about it, People who use Image Data and Video data could also be part of surveillance teams hence educating them about Model Bias could help a socially unbiased AI s olution.
Kaggle Learn, Coursera and Udemy need to improve their Course curriculum to include lessons about Model Bias and Fairness.
Fast.AI, TheSchool.Ai need to have courses focused on Interpretable Machine Learning.
To improve Reprodcibility in Machine Learning, it is essential to introduce tools that are as easy as uploading code on github but also that can have a completely reproducible project
Coding Curriculum encouraging and training new comers using tools like Rmarkdown can help in better documentation
Training on Docker and Virtual Box and encouraing to use Kaggle Kernels/Collabs should drive more reproducible work
Ultimately, the Incentive of delivering a reproducible work should be applauded over non-reproducible works. In other words, non-reproducilble work must require extra scrutiny or lower-tier while reproducible work should have autoapproval work and verified tags as a sign of better incentive.

Media Coverage about Bias / IML

Amazon scraps secret AI recruiting tool that showed bias against women

Undoing Bias / Improving IML - Papers and Attempts

ML Bias, Interpretability - Kaggler Perspective & Recommendations

Do Computers Lie?

Machine Bias

The Problem

The Solution

Libraries

Ignorance is Bliss - but not always!

Reproducibility, Explainability and Bias

Model Bias & Model Fairness

Who are they?

Gender

Age

Student vs Professionals

Undergraduate Majors

R vs Python vs more

All Countries

Countries with at least 100 respondents

Media Sources

Work Industry

Data Scientist?

Type of Data

MOOC - Online course Platform

Percentage of Data Projects

Metrics for Model Success

Difficulty in implementing a Fair / Bias-free ML Model

Model Interpretability / Explaining ML Models

Who are they?

Gender

Age

R vs Python vs more

Degree - Education

Work Industry

MOOC - Online course Platform

Media Sources

Exploring model insights? - Percentage of Data Projects

ML models to be black boxes - Difficult to explain

Circumstances where exploring model insights and interpreting model happens

Preferred Methods for Model Interpretability

Reproducibility

Tools and Methods for ML Reproduction

Barriers preventing from making easier to reuse and reproducible work

Recommendations from these Insights

Media Coverage about Bias / IML

Undoing Bias / Improving IML - Papers and Attempts